Pretty good, pretty , prittty, pritttty good...

Discussion of anything and everything relating to chess playing software and machines.

Moderators: hgm, Harvey Williamson, bob

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
User avatar
Ovyron
Posts: 2325
Joined: Tue Jul 03, 2007 2:30 am

Re: Pretty good, pretty , prittty, pritttty good...

Post by Ovyron » Wed Jan 31, 2018 10:37 pm

mroh wrote:What do you do if the refutation of a line is a move of the opposite side? (which sounds pretty natural to me)
Huh? You play it on the board and let Stockfish see it so it writes it to the Hash and you can go back to some previous position with this information on the hash so Stockfish shows a different continuation?

Refutations of lines come from both sides and what the user does is no different, what matters is the engine sees its plan doesn't work (from either side) so it can suggest a different path.

Anyway, the hash is unreliable, it'll be overwritten, it's better to just feed everything to an engine with Persistent Analysis to store all the searches or relying on a GUI to store a tree with all relevant positions backsolved.

What a default Contempt that is not zero has shown is that Stockfish had actually become two different engines. One plays great the white side but lousy the black side (say, what'd you get with negative Contempt), the other plays great the black side but lousy the white side.

Contempt = 0 of CFish looks like a compromise, the middle ground.

It's clear to me that for analysis you want the one that plays greatly the side you're analyzing. But there's two different Stockfishes, so it doesn't make sense to pollute the analysis of the great white player with the one from the lousy one (Stockfish plays bad the white side when analyzing the black side), so while it might be impractical to use two instances of Stockfish, one for each side, it seems it's the best way to get these great moves that Contempt=0 misses.

mroh
Posts: 29
Joined: Thu Nov 02, 2017 5:51 am
Location: Germany
Contact:

Re: Pretty good, pretty , prittty, pritttty good...

Post by mroh » Wed Jan 31, 2018 11:25 pm

Ok, now I get it, I think. (Sorry for being slow ^^)

The only thing the gui would need to take care of is to persist the result of this analysis with two different "views", one for each side, so they dont get polluted, as you say.
(Its nearly the same as the "human view", we talked about... I could even save the trees for several different Contempts...)

Its an interesting concept, for sure!

Think, I need to make same experience with it...

User avatar
Ovyron
Posts: 2325
Joined: Tue Jul 03, 2007 2:30 am

Re: Pretty good, pretty , prittty, pritttty good...

Post by Ovyron » Thu Feb 01, 2018 12:06 am

mroh wrote:The only thing the gui would need to take care of is to persist the result of this analysis with two different "views", one for each side, so they dont get polluted, as you say.
Indeed, no GUI currently supports this new paradigm.

Fortunately, my GUI has a comment box which allows me to paste in it whatever I want, like Window's Wordpad, so I've been able to store the engine's backsolved scores, and I backsolve them manually with two different "views".

I use colored numbers to save time, instead of having to type the engines' name, my starting position looks like this:

Image

But I haven't backsolved to the root of the starting position for a while, an engine with Persistent Analysis is just more efficient.

Post Reply